thinking level variants

[JayGhiya]

For some models xhigh thinking is supported can we get that in?

[sidkang]

i just setup a recommended thinking level in the AGENTS.md for each agent, it works well for now.